Method and apparatus for wind noise detection and suppression using multiple microphones

• US 8,924,204 B2
• Filed: 09/30/2011
• Issued: 12/30/2014
• Est. Priority Date: 11/12/2010
• Status: Active Grant

• Current Assignee: Avago Technologies International Sales Pte Limited (Broadcom, Inc.)
• Original Assignee: Broadcom Corporation (Broadcom, Inc.)
• Inventors: Chen, Juin-Hwey, Thyssen, Jes, Zhang, Xianxian, Zeng, Huaiyu
• Primary Examiner(s): Shah, Paras D
• Assistant Examiner(s): THOMAS-HOMESCU, ANNE L
• Application Number:
  US13/250,291
• Publication Number:
  US 20120123771A1
• Time in Patent Office: 1,187 Days
• Field of Search: 704/226, 704/233, 704/262, 704/211, 704/210, 704/207, 704/214, 704/500, 704/205, 704/268, 704/239, 704/219, 381/313, 381/92, 381/71.1, 381 941- 942, 381/122, 381/60, 381/309, 381/109, 381/22, 381 56- 57, 381/107, 455/501, 455/67.11, 455/456.1, 455/556.1, 455/115.1
• US Class Current: 704/226
• CPC Class Codes: G10L 2021/02165 Two microphones, one receiving mainly the noise signal and the other one mainly the speech signal
• CPC Class Codes: G10L 21/0208 Noise filtering
• CPC Class Codes: G10L 21/0272 Voice signal separating
• CPC Class Codes: H04R 1/245 of microphones
• CPC Class Codes: H04R 2410/07 Mechanical or electrical reduction of wind noise generated by wind passing a microphone

Abstract

Unlike sound based pressure waves that go everywhere, air turbulence caused by wind is usually a fairly local event. Therefore, in a system that utilizes two or more spatially separated microphones to pick up sound signals (e.g., speech), wind noise picked up by one of the microphones often will not be picked up (or at least not to the same extent) by the other microphone(s). Embodiments of methods and apparatuses that utilize this fact and others to effectively detect and suppress wind noise using multiple microphones that are spatially separated are described.

Claims

1. An apparatus for detecting and suppressing wind noise in a primary signal received by a primary microphone, the apparatus comprising:

a reference signal adjustment processor, in communication with a memory, configured to adjust a reference signal received by a reference microphone based on a difference between a parameter of the reference signal and a parameter of the primary signal to provide an adjusted reference signal, wherein the parameter of the reference signal and the parameter of the primary signal is a delay, gain, spectral shape, or background noise level, wherein the reference signal is adjusted to better match the parameter of the reference signal to the parameter of the primary signal; and

a waveform substitution processor, in communication with a memory, configured to substitute at least a portion of a frame of samples of the primary signal with at least a portion of a frame of samples of the adjusted reference signal if a primary microphone wind noise detection signal indicates that wind noise is present or above a first threshold in the frame of samples of the primary signal and a reference microphone wind noise detection signal indicates the wind noise is absent or below a second threshold in the frame of samples of the adjusted reference signal.

2. The apparatus of claim 1, wherein the waveform substitution processor is further configured to smooth waveform discontinuities between the primary microphone signal and the frame of samples of the adjusted reference signal by performing an overlap-add operation.

3. The apparatus of claim 1, further comprising:

a single-channel noise suppression processor, in communication with a memory, configured to provide a reduction of acoustic noise in the frame of samples of the primary signal by a particular amount if the primary microphone wind noise detection signal indicates that wind noise is absent or below a third threshold in the frame of samples of the primary signal and the reference microphone wind noise detection signal indicates the wind noise is present or above a fourth threshold in the frame of samples of the adjusted reference signal.

4. The apparatus of claim 3, wherein the particular amount is determined based on an average amount of acoustic noise reduction in the primary signal provided by a multi-channel noise suppression processor.

5. The apparatus of claim 1, further comprising:

a packet loss concealment processor, in communication with the memory, configured, to replace the frame of samples of the primary signal with samples extrapolated from previous samples of the primary signal if the primary microphone wind noise detection signal indicates that wind noise is present or above a third threshold in the frame of samples of the primary signal and the reference microphone wind noise detection signal indicates the wind noise is present or above a fourth threshold in the frame of samples of the adjusted reference signal.

6. The apparatus of claim 5, wherein the packet loss concealment processor is further configured to generate the samples extrapolated from previous samples of the primary signal using speech parameters estimated from the frame of samples of the primary signal or the frame of samples of the adjusted reference signal.

7. The apparatus of claim 1, further comprising:

a weighted sum processor, in communication with a memory, configured to replace the frame of samples of the primary signal with a weighted sum of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal if the primary microphone wind noise detection signal indicates that wind noise is present or above a third threshold in the frame of samples of the primary signal and the reference microphone wind noise detection signal indicates the wind noise is present or above a fourth threshold in the frame of samples of the adjusted reference signal.

8. The apparatus of claim 7, wherein the weighted sum processor is further configured to weight the frame of samples of the primary signal and the frame of samples of the adjusted reference signal based on a ratio of an estimated wind noise energy in the frame of samples of the primary signal to an estimated wind noise energy in the frame of samples of the adjusted reference signal.

9. The apparatus of claim 1, wherein the reference signal adjustment processor comprises:

an adaptive filter configured to adjust the reference signal based on the difference between the parameter of the reference signal and the parameter of the primary signal to provide the adjusted reference signal.

10. The apparatus of claim 9, wherein the reference signal adjustment processor is further configured to derive tap coefficients of the adaptive filter from tap coefficients of an additional adaptive filter configured to filter the primary signal to approximate a speech component in the reference signal.

11. The apparatus of claim 1, further comprising:

a multi-method wind noise detection processor, in communication with the memory, configured to generate first and second wind noise detection signals that indicate whether wind noise is present or absent in the frame of samples of the primary signal; and

a wind noise detection signal combining processor, in communication with the memory, configured to combine the first and second wind noise detection signals to provide the primary microphone wind noise detection signal.

12. The apparatus of claim 11, wherein the multi-method wind noise detection processor comprises:

a correlation based wind noise detection processor, in communication with a memory, configured to generate the first wind noise detection signal based on;

a correlation of the frame of samples of the primary signal with a frame of samples of the reference signal,

a correlation of the frame of samples of the primary signal with a second frame of samples of the primary signal, wherein the second frame of samples of the primary signal are in an estimated pitch period range of the frame of samples of the primary signal, and

a correlation of the frame of samples of the reference signal with a second frame of samples of the reference signal, wherein the second frame of samples of the reference signal are in an estimated pitch period range of the frame of samples of the reference signal.

13. The apparatus of claim 11, wherein the multi-method wind noise detection processor comprises:

an average gain difference based wind noise detection processor, in communication with a memory, configured to generate the first wind noise detection signal based on an average difference between corresponding frequency component magnitudes of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal.

14. The apparatus of claim 13, wherein the corresponding frequency component magnitudes of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal are expressed according to a logarithmic scale.

15. The apparatus of claim 13, wherein the corresponding frequency component magnitudes of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal are limited to frequency component magnitudes associated with frequencies in a range determined based on a frequency spectrum associated with wind noise.

16. The apparatus of claim 11, wherein the multi-method wind noise detection processor comprises:

a signal-to-matching-noise ratio based wind noise detection processor, in communication with a memory, configured to generate the first wind noise detection signal based on an energy ratio of the frame of samples of the primary signal to a difference between the frame of samples of the primary signal and the frame of samples of the adjusted reference signal.

17. A method for detecting and suppressing wind noise in a primary signal received by a primary microphone. the method comprising:

adjusting a reference signal received by a reference microphone based on a difference between a parameter of the reference signal and a parameter of the primary signal to provide an adjusted reference signal, wherein the parameter of the reference signal and the parameter of the primary signal is a delay, gain, spectral shape, or background noise level, wherein the reference signal is adjusted to better match the parameter of the reference signal to the parameter of the primary signal; and

substituting at least a portion of a frame of samples of the primary signal with at least a portion of a frame of samples of the adjusted reference signal if a primary microphone wind noise detection signal indicates that wind noise is present or above a first threshold in the frame of samples of the primary signal and a reference microphone wind noise detection signal indicates the wind noise is absent or below a second threshold in the frame of samples of the adjusted reference signal.

18. The method of claim 17, wherein substituting the at least the portion of the frame of samples of the primary signal with the at least the portion of the frame of samples of the adjusted reference signal further comprises:

smoothing waveform discontinuities between the primary signal and the frame of samples of the adjusted reference signal by performing an overlap-add operation.

19. The method of claim 17, further comprising:

performing single-channel noise suppression to reduce acoustic noise in the frame of samples of the primary signal by a particular amount if the primary microphone wind noise detection signal indicates that wind noise is absent or below a third threshold in the frame of samples of the primary signal and the reference microphone wind noise detection signal indicates the wind noise is present or above a fourth threshold in the frame of samples of the adjusted reference signal.

20. The method of claim 19, wherein the particular amount is determined based on an average amount of acoustic noise reduction in the primary signal provided by a multi-channel noise suppression module.

21. The method of claim 17, further comprising:

replacing the frame of samples of the primary signal with samples extrapolated from previous samples of the primary signal if the primary microphone wind noise detection signal indicates that wind noise is present or above a third threshold in the frame of samples of the primary signal and the reference microphone wind noise detection signal indicates the wind noise is present or above a fourth threshold in the frame of samples of the adjusted reference signal.

22. The method of claim 21, wherein replacing the frame of samples of the primary signal with samples extrapolated from previous samples of the primary signal further comprises:

generating the samples extrapolated from the previous samples of the primary signal using speech parameters estimated from the frame of samples of the primary signal or the frame of samples of the adjusted reference signal.

23. The method of claim 17, further comprising:

replacing the frame of samples of the primary signal with a weighted sum of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal if the primary microphone wind noise detection signal indicates that wind noise is present or above a third threshold in the frame of samples of the primary signal and the reference microphone wind noise detection signal indicates the wind noise is present or above a fourth threshold in the frame of samples of the reference signal.

24. The method of claim 23, wherein replacing the frame of samples of the primary signal with the weighted sum of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal comprises:

weighting the frame of samples of the primary signal and the frame of samples of the adjusted reference signal based on a ratio of an estimated wind noise energy in the frame of samples of the primary signal to an estimated wind noise energy in the frame of samples of the adjusted reference signal.

25. The method of claim 17, wherein the adjusting the reference signal received by the reference microphone is performed, at least in part, by an adaptive filter configured to adjust the reference signal based on the difference between the parameter of the reference signal and the parameter of the primary signal to provide the adjusted reference signal.

26. The method claim 25, further comprising:

deriving tap coefficients of the adaptive filter from tap coefficients of an additional adaptive filter configured to filter the primary signal to approximate a speech component in the reference signal.

27. The method of claim 17, further comprising:

generating first and second wind noise detection signals that indicate whether wind noise is present or absent in the frame of samples of the primary signal; and

combining the first and second wind noise detection signals to provide the primary microphone wind noise detection signal.

28. The method of claim 27, wherein generating the first wind noise detection signal comprises:

correlating the frame of samples of the primary signal with the frame of samples of the reference signal;

correlating the frame of samples of the primary signal with a second frame of samples of the primary signal, wherein the second frame of samples of the primary signal are in an estimated pitch period range of the frame of samples of the primary signal; and

correlating the frame of samples of the reference signal with a second frame of samples of the reference signal, wherein the second frame of samples of the reference signal are in an estimated pitch period range of the frame of samples of the reference signal.

29. The method of claim 27, wherein generating the first wind noise detection signal comprises:

determining an average difference between corresponding frequency component magnitudes of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal.

30. The method of claim 29, wherein the corresponding frequency component magnitudes of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal are expressed according to a logarithmic scale.

31. The method of claim 29, wherein the corresponding frequency component magnitudes of the frame of samples of the primary signal and the frame of samples of the adjusted reference signal are limited to frequency component magnitudes associated with frequencies in a range determined based on a frequency spectrum associated with wind noise.

32. The method of claim 27, wherein generating the first wind noise detection signal comprises:

determining an energy ratio of the frame of samples of the primary signal to a difference between the frame of samples of the primary signal and a corresponding frame of samples of the adjusted reference signal.